Amazon Web Services has upgraded throughput and I/O on Elastic Block Store volumes backed by solid-state drives to allow customers to migrate highly performance-sensitive database workloads into the cloud.
The move comes two months after Amazon Web Services (AWS) made Elastic Block Store (EBS) volumes backed by solid-state drives (SSDs) the default option for block storage. Now, the maximum sustained throughput to SSD-backed AWS EBS volumes has doubled to 128 MB per second, and disk I/O will also get a boost with a new 256 KB block size, up from 16 KB, according to an AWS blog post.
"We don't have a large number of implementations using [provisioned IOPS] volumes but I would imagine this … would allow us to implement fewer PIOPS volumes than before," said James Fogerson, solutions architect for international staffing firm Robert Half International Inc., in Menlo Park, Calif. "I think the overall improved performance volumes will benefit us, especially as we implement more big data workloads to AWS."
A major barrier to cloud adoption for high-performance workloads has always been storage performance, said Jared Reimer, co-founder of Cascadeo Corp., an IT consulting firm located in Mercer Island, Wash.
"Big database workloads that have high transaction rates or don't cache well … either were infeasible entirely or were cost-prohibitive in the cloud, and now that's changing," Reimer said.
One of Reimer's clients has considered building a new data center or migrating to Amazon, and this will push them further toward the cloud, he said.
Adoption has been high for SSD-backed AWS EBS general purpose and provisioned IOPS volumes, according to Amazon's blog post.
"The fact that there's been so much uptake on it validates [that] if I have this problem, other people have it too," Reimer said. "It tells me that there's a lot of pent-up demand and a lot of companies have struggled with this."
Larger EBS volumes and advanced availability wanted
Performance upgrades to AWS EBS volumes are welcome, but a boost in the maximum volume size offered on EBS would also be desirable for some customers.
Today, the volume size limit is one terabyte (TB), and volumes can be aggregated to support arbitrarily large logical volumes, but customers such as Fogerson would rather avoid the complexity and abstraction of doing that.
Jared Reimer co-founder of Cascadeo Corp.
"If we were to implement a SharePoint project, we would talk about storing 4 TB of data and most of the effort would be on getting the data to AWS and how to store it and back-it up," Fogerson said. "Larger EBS volumes would be nice so we don't have to play around with multiple volumes."
EBS volumes are good for clustered database workloads, Reimer said, but support for advanced high-availability techniques would further sweeten the deal for performance-sensitive database users.
AWS currently offers support for master-slave database replication, read replicas, or managed databases through the Relational Database Service. But multi-master clusters that perform reads and writes concurrently to the same volumes aren't uncommon among high-performance workloads today, Reimer said.
"Anything they can do to improve the ability of end users to create highly available database clusters is significant," Reimer said. "As they keep working toward enterprise and governmental adoption the number of customers that are going to demand that kind of availability and scalability in database storage is going to increase."